Mobile Computer-Mediated Assessment of Autism Risk By Non-Specialists in Home Settings: Insights from the START Project

Background: In many parts of the world with poor access to clinical expertise, a large number of children on the autism spectrum remain undetected, and hence are deprived of early interventions. Even those who are diagnosed experience significant delays between parents’ noticing atypicalities in behaviour and specialist clinical diagnosis. To bridge this detection gap, we have developed START (“Screening Tool for Autism Risk using Technology”), a tablet-based app usable by minimally-trained non-specialist workers in home settings. START presents a range of tasks as simple games for an evaluation of social, motor, sensory, and cognitive function in 2-7-year-old children.

Objectives: To conduct a proof-of-principle study with START, administered by non-specialist workers in and around Delhi, India, using a case control study design comprising three groups of children: those with autism spectrum disorders (ASD), intellectual disabilities (ID), or typically developing (TD) children.

Methods: Non-specialist workers tested 127 children with ASD (n=46), ID (n=38) and TD (n=43), aged 2-7 years in their homes. We present data on three sets of measures: (1) social motivation measured using a) preferential looking task: tablet-based eye-tracking to measure distribution of overt attention between a social and a non-social stimulus, and b) choice task: to measure preference for social rewards; (2) sensory interests were measured by showing children a video of a spinning wheel which they could stop at will; and (3) motor following task: where participants were asked to follow with their finger the trajectory of a butterfly moving in a predetermined random manner.

Results: (1) In the preferential gaze task, children with ASD spend significantly (Figure 1, top left) less time looking at the social stimuli than the TD group (F(2, 104) = 4.21, p = 0.017) but in the choice task (Figure 1, top right) the groups do not differ in their choice of the social stimulus (F(2, 98) = 0.58, ns). (2) Children with ASD spend significantly longer looking at a spinning wheel than TD (F(2, 81) = 19.57, p < .001). (3) Children with ASD execute the motor following task with significantly greater spatio-temporal mean squared errors than do TD or ID groups (F(2, 78) = 19.36, p < .001).

Conclusions: This study provides proof of principle for START, revealing the expected pattern of ASD-TD group differences in relevant tasks. It also demonstrates the feasibility of such mobile scalable testing of neurodevelopment by non-specialists in home settings.